Hands-On Large Language Models

AI 語言技術的進步，正以前所未有的速度改變世界！「Jay和Maarten一如既往地運用精美的插圖為複雜主題提供深具洞察的描述。對於任何想了解大型語言模型背後主要技術的人來說，這本書都是寶貴的資源。」—Andrew Ng（吳恩達)，DeepLearning.AI 創辦人「我想不出有哪本書比這本書更重要，值得現在就去閱讀。書中的每一頁，我都學到在這個語言模型時代，對於成功至關重要的知識。」—Josh Starmer，StatQuest想深入了解大型語言模型的奧秘，並將其應用於實際專案中？《Hands-On Large Language Models》正是為你量身打造的指南！由知名AI專家Jay Alammar和Maarten Grootendorst執筆，透過近300幅精美圖解，帶你輕鬆掌握LLM的核心概念和實踐技巧。無論是文案撰寫、摘要生成，還是建立先進的語義搜尋系統，本書都提供了詳盡的步驟和範例，助你在AI領域中脫穎而出。快讓《Hands-On Large Language Models》開啟你的 AI 進階之旅！--------------------------------------------------------近年來，AI在語言處理領域取得了驚人的突破。隨著深度學習的快速發展，語言AI系統在文字寫作與理解方面的能力，比以往任何時候都更加出色。這一趨勢正在催生全新的功能、產品，甚至是全新的產業。透過本書視覺化的教學方式，你將學習如何運用這些強大技術，掌握實用工具與關鍵概念，並能立即應用於你的專案中！你將學會如何運用預訓練的 LLM 進行文案撰寫與摘要生成；打造超越傳統關鍵字匹配技術的語義搜尋系統；運用現有的函式庫和預訓練模型，來進行文本分類、搜尋與聚類。本書將助你深入理解：• Transformer語言模型的架構，掌握文本生成與語意表達的核心技術。• 構建高效的LLM流程，對文本資料進行聚類並探索其主題。• 透過密集檢索（dense retrieval）與重新排名（reranking），構建超越關鍵字匹配的語義搜尋引擎。• 從提示工程（prompt engineering）到檢索擴增生成（RAG），探索生成式AI的各種運用方式。• 深入學習如何訓練 LLM，針對特定應用進行優化，包括生成式微調（generative fine-tuning）、對比微調（

This book provides an overview of multiple aspects of spatial language understanding and explores recent trends of modeling based on very large foundational models and their applications. The authors address the following five main themes: spatial semantic representations in both symbolic and sub-symbolic spaces; spatial information extraction computational models; reasoning over spatial language; commonsense spatial understanding; and multiple modern and state-of-the-art downstream applications of spatial language understanding including dialogue systems, narrative discourse, and grounding language in the physical world with the multi-modal problem settings. One of the essential functions of natural language is to express spatial relationships between objects. Linguistic constructs can encode highly complex structures of objects, spatial relations between them, and patterns of motion through space relative to a reference point. The complexity of spatial language understanding and its

Every day we interact with machine learning systems offering individualized predictions for our entertainment, social connections, purchases, or health. These involve several modalities of data, from sequences of clicks to text, images, and social interactions. This book introduces common principles and methods that underpin the design of personalized predictive models for a variety of settings and modalities. The book begins by revising 'traditional' machine learning models, focusing on adapting them to settings involving user data, then presents techniques based on advanced principles such as matrix factorization, deep learning, and generative modeling, and concludes with a detailed study of the consequences and risks of deploying personalized predictive systems. A series of case studies in domains ranging from e-commerce to health plus hands-on projects and code examples will give readers understanding and experience with large-scale real-world datasets and the ability to design mod

Downscaling is a widely used technique for translating information from large-scale climate models to the spatial and temporal scales needed to assess local and regional climate impacts, vulnerability, risk and resilience. This book is a comprehensive guide to the downscaling techniques used for climate data. A general introduction of the science of climate modeling is followed by a discussion of techniques, models and methodologies used for producing downscaled projections, and the advantages, disadvantages and uncertainties of each. The book provides detailed information on dynamic and statistical downscaling techniques in non-technical language, as well as recommendations for selecting suitable downscaled datasets for different applications. The use of downscaled climate data in national and international assessments is also discussed using global examples. This is a practical guide for graduate students and researchers working on climate impacts and adaptation, as well as for polic

The plural form 'Englishes' conveys the diversity of English as a global language, pinpointing the growth and existence of a large number of national, regional and social forms. The global spread of English and the new varieties that have emerged around the world has grown to be a vast area of study and research, which intersects multiple disciplines. This Handbook provides a comprehensive and authoritative survey of World Englishes from 1600 to the present day. Covering topics such as variationist sociolinguistics, pragmatics, contact linguistics, linguistic anthropology, corpus- and applied linguistics and language history, it combines discussion of traditional topics with a variety of innovative approaches. The chapters, all written by internationally acclaimed authorities, provide up-to-date discussions of the evolution of different Englishes around the globe, a comprehensive coverage of different models and approaches, and some original perspectives on current challenges.

A large body of knowledge has accumulated in recent years on the cognitive processes underlying language, much of which comes from studies of Indo-European languages, in particular English. This groundbreaking volume explores the languages of South and Southeast Asia, which differ significantly from Indo-European languages in their grammar, lexicon and spoken forms. This book raises new questions in psycholinguistics and enables readers to re-evaluate previous models in light of new research. With thirty-six chapters divided into three parts - Language Acquisition, Language Processing and Language and Brain - it examines contemporary topics alongside new findings in areas such as first and second language acquisition, the development of literacy, the diagnosis of language and reading disorders, and the relationship between language, brain, culture and cognition. It will be invaluable to all those interested in the languages of South and Southeast Asia, as well as psychologists, linguis

A solid introduction to mathematical modeling for a range of chemical engineering applications, covering model formulation, simplification and validation. It explains how to describe a physical/chemical reality in mathematical language and how to select the type and degree of sophistication for a model. Model reduction and approximation methods are presented, including dimensional analysis, time constant analysis and asymptotic methods. An overview of solution methods for typical classes of models is given. As final steps in model building, parameter estimation and model validation and assessment are discussed. The reader is given hands-on experience of formulating new models, reducing the models and validating the models. The authors assume the knowledge of basic chemical engineering, in particular transport phenomena, as well as basic mathematics, statistics and programming. The accompanying problems, tutorials, and projects include model formulation at different levels, analysis, pa

A large body of knowledge has accumulated in recent years on the cognitive processes underlying language, much of which comes from studies of Indo-European languages, in particular English. This groundbreaking volume explores the languages of South and Southeast Asia, which differ significantly from Indo-European languages in their grammar, lexicon and spoken forms. This book raises new questions in psycholinguistics and enables readers to re-evaluate previous models in light of new research. With thirty-six chapters divided into three parts - Language Acquisition, Language Processing and Language and Brain - it examines contemporary topics alongside new findings in areas such as first and second language acquisition, the development of literacy, the diagnosis of language and reading disorders, and the relationship between language, brain, culture and cognition. It will be invaluable to all those interested in the languages of South and Southeast Asia, as well as psychologists, linguis

Hands-on, theory-neutral and non-technical, this textbook is a basic introduction to the structure of English words and sentences. Assuming no prior knowledge of linguistic analysis, it presents the facts in a straightforward manner and offers a step-by-step guide from small to large building blocks of language. Every chapter contains numerous exercises and discussion questions, which provide essential self-study material, as well as in-chapter tasks which lead students to a more comprehensive understanding of linguistic issues. The book also features concise chapter summaries, suggestions for further reading, an inclusive glossary and two consolidation chapters which encourage students to secure their understanding of the English language. The dedicated companion website includes further exercises, answers and solutions to the exercises, as well as useful links.

Covering point process theory, random geometric graphs and coverage processes, this rigorous introduction to stochastic geometry will enable you to obtain powerful, general estimates and bounds of wireless network performance and make good design choices for future wireless architectures and protocols that efficiently manage interference effects. Practical engineering applications are integrated with mathematical theory, with an understanding of probability the only prerequisite. At the same time, stochastic geometry is connected to percolation theory and the theory of random geometric graphs and accompanied by a brief introduction to the R statistical computing language. Combining theory and hands-on analytical techniques with practical examples and exercises, this is a comprehensive guide to the spatial stochastic models essential for modelling and analysis of wireless network performance.

This advanced textbook on modeling, data analysis and numerical techniques for marine science has been developed from a course taught by the authors for many years at the Woods Hole Oceanographic Institute. The first part covers statistics: singular value decomposition, error propagation, least squares regression, principal component analysis, time series analysis and objective interpolation. The second part deals with modeling techniques: finite differences, stability analysis and optimization. The third part describes case studies of actual ocean models of ever increasing dimensionality and complexity, starting with zero-dimensional models and finishing with three-dimensional general circulation models. Throughout the book hands-on computational examples are introduced using the MATLAB programming language and the principles of scientific visualization are emphasised. Ideal as a textbook for advanced students of oceanography on courses in data analysis and numerical modeling, the boo

Lexical semantics has become a major research area within computational linguistics, drawing from psycholinguistics, knowledge representation, computer algorithms and architecture. Research programmes whose goal is the definition of large lexicons are asking what the appropriate representation structure is for different facets of lexical information. Among these facets, semantic information is probably the most complex and the least explored. Computational Lexical Semantics is one of the first volumes to provide models for the creation of various kinds of computerized lexicons for the automatic treatment of natural language, with applications to machine translation, automatic indexing, and database front-ends, knowledge extraction, among other things. It focuses on semantic issues, as seen by linguists, psychologists and computer scientists. Besides describing academic research, it also covers ongoing industrial projects.

This book offers a unique insight into a revolution in software development that allows model specifications to be fully and efficiently translated into code. Using the most widely adopted, industry standard, software modelling language, UML, the reader will learn how to build robust specifications based on OMG's Model Driven Architecture (MDA). From there, the authors describe the steps needed to translate the Executable UML (xUML) models to any platform-specific implementation. The benefits of this approach go well beyond simply reducing or eliminating the coding stage - it also ensures platform independence, avoids obsolescence (programming languages may change, the model doesn't) and allows full verification of the models by executing them in a test and debug xUML environment. This is an excellent reference for anyone embarking on what is surely the future of software development for medium and large scale projects.